Chasing Real Impact with GenAI in Healthcare:
Part 1 – Trends and Doctor Perspectives
April 24th, 2025 | Klemen Vodopivec, Nejc Završan
Introduction
In a recent blog post, we explored how patients are beginning to use large language models (LLMs) like ChatGPT in their healthcare journeys. But if consumers are turning to these tools, the obvious follow-up is: Are doctors using them, too?
The answer is yes.
Healthcare today faces immense challenges, and there’s a growing belief — one we share at MedAsk — that generative AI (GenAI) can help address them. But the real challenge isn’t just building an LLM product for healthcare; it’s understanding exactly how and where it can be practically integrated without becoming an additional burden.
In this two-part blog series, we’ll share our work exploring practical applications of LLMs in healthcare.
- In Part 1 (this post), we examine broader industry trends and share our insights from interviews with seven doctors across five private family clinics.
In Part 2, we’ll dive into a focused case study of MedAsk at one of those clinics, including patient feedback gathered through a questionnaire.
Exploring GenAI Adoption in Healthcare
How Doctors Are Already Embracing GenAI in Practice
Despite limited guidance and unclear workplace policies, many general practitioners (GPs) are already using GenAI to support their work. According to a study published in BMJ, one in five UK GPs surveyed reported using LLM-powered chatbots to assist in clinical practice. Among those who responded, 29% said they use LLMs to generate documentation after patient appointments, while 28% use them to suggest differential diagnoses.
Another recent survey by Fierce Healthcare provided one of the first detailed examinations of physician use of publicly available GenAI tools. Among over 100 physicians surveyed, over half used them to support diagnosis, nearly half for generating clinical documentation, and more than 40% for treatment planning. Additionally, an impressive 70% said they use these tools for patient education and literature searches. These trends, highlighted in Figure 1, underscore how LLMs are quickly becoming integrated into clinicians’ day-to-day routines. As this adoption continues, attention is increasingly turning to how GenAI might drive broader transformations across healthcare.
Triage is Becoming a Key Use Case for GenAI in Healthcare
There is growing interest in reimagining how triage and patient navigation work in healthcare. The current systems are often fragmented, inconsistent, and labor-intensive. An AI-powered assistant has the potential to transform this process by automating repetitive tasks, improving staff productivity, and enabling new workflows that were previously impossible within time or resource constraints.
Rule-based tools, like symptom checkers, have shown some promise in this area. They’ve helped clinics by providing preliminary patient information and reducing call volumes. For patients, they offer quicker access to actionable information about their health. But in many cases, they’ve often created extra work: clinicians need to follow up with phone calls to make sense of inconsistent or inaccurate results.
We believe generative AI represents a step-change improvement over these rigid rule-based systems, which lack the flexibility and nuance needed for effective triage. Our comparative testing has shown MedAsk’s potential to significantly outperform traditional tools in both diagnostic and triage accuracy.
But to clinicians, benchmark numbers alone don’t matter. What matters is whether the tool is usable and can be embedded into real workflows without adding friction. That’s why much of our focus is on gathering feedback from both doctors and patients in real clinical settings.
What Doctors Told Us About MedAsk’s Role in Clinical Workflows
We conducted unstructured interviews with seven doctors across five private family medicine clinics. From these conversations, we identified several practical ways MedAsk could improve clinical workflows:
- Triage support: Patients can use MedAsk to self-book appointments rather than calling or emailing the clinic. This improves access, reduces administrative burden, and helps clinics manage demand more effectively.
- E-consultation relief: Doctors noted a growing number of e-consultations, many of which take a lot of time to solve (even though they are unpaid) and involve repetitive symptom-related questions. MedAsk can handle a significant portion of these, freeing up capacity for more meaningful, in-person care.
- Better symptom detail from patients: MedAsk encourages patients to share more thorough and structured symptom descriptions. This reduces the back-and-forth often needed to clarify a condition, especially in digital channels.
- Pre-visit reports: The tool generates a concise summary of the patient’s interaction before the appointment, helping doctors come in better prepared.
- Anamnesis templates: Doctors can use the pre-visit report as a starting point for writing the required anamnesis (medical history) after the consultation, saving time and effort by editing rather than starting from scratch.
- Reference aid: In cases where a diagnosis was uncertain, doctors appreciated having access to MedAsk’s list of potential conditions and reasoning as a reference.
Conclusion
GenAI adoption among doctors is accelerating. And while healthcare has traditionally been slow to implement new technologies, we’re now seeing strong signals—especially in areas like digital triage, diagnostic support, and reducing administrative overhead. Digital triage, in particular, stood out as a promising use case in our conversations with doctors. It reduces the load on phone lines and email inboxes, giving doctors more time to focus on patient care. It also helps both patients and doctors come into appointments better prepared and saves time post-visit by supporting documentation like anamnesis.
At MedAsk, we believe the path to real impact isn’t paved by theory alone. It’s built through fieldwork: talking to doctors and patients, observing real workflows, and testing AI tools in safe, practical settings.
In Part 2 of this series, we’ll share findings from our patient-side case study at one of the clinics, including first hand feedback from those using MedAsk directly.